2 * Test program for MSA instruction DOTP_U.W
4 * Copyright (C) 2019 Wave Computing, Inc.
5 * Copyright (C) 2019 Aleksandar Markovic <amarkovic@wavecomp.com>
6 * Copyright (C) 2019 RT-RK Computer Based Systems LLC
7 * Copyright (C) 2019 Mateja Marjanovic <mateja.marjanovic@rt-rk.com>
9 * This program is free software: you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation, either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program. If not, see <https://www.gnu.org/licenses/>.
27 #include "../../../../include/wrappers_msa.h"
28 #include "../../../../include/test_inputs_128.h"
29 #include "../../../../include/test_utils_128.h"
31 #define TEST_COUNT_TOTAL ( \
32 (PATTERN_INPUTS_SHORT_COUNT) * (PATTERN_INPUTS_SHORT_COUNT) + \
33 (RANDOM_INPUTS_SHORT_COUNT) * (RANDOM_INPUTS_SHORT_COUNT))
38 char *instruction_name
= "DOTP_U.W";
41 struct timeval start
, end
;
44 uint64_t b128_result
[TEST_COUNT_TOTAL
][2];
45 uint64_t b128_expect
[TEST_COUNT_TOTAL
][2] = {
46 { 0xfffc0002fffc0002ULL
, 0xfffc0002fffc0002ULL
, }, /* 0 */
47 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
48 { 0x5552aaac5552aaacULL
, 0x5552aaac5552aaacULL
, },
49 { 0xaaa95556aaa95556ULL
, 0xaaa95556aaa95556ULL
, },
50 { 0x9996666899966668ULL
, 0x9996666899966668ULL
, },
51 { 0x6665999a6665999aULL
, 0x6665999a6665999aULL
, },
52 { 0x1c6fe38f71c48e3aULL
, 0xc71a38e51c6fe38fULL
, },
53 { 0xe38c1c738e3771c8ULL
, 0x38e1c71de38c1c73ULL
, },
54 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, }, /* 8 */
55 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
56 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
57 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
58 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
59 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
60 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
61 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
62 { 0x5552aaac5552aaacULL
, 0x5552aaac5552aaacULL
, }, /* 16 */
63 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
64 { 0xe38c71c8e38c71c8ULL
, 0xe38c71c8e38c71c8ULL
, },
65 { 0x71c638e471c638e4ULL
, 0x71c638e471c638e4ULL
, },
66 { 0x110eeef0110eeef0ULL
, 0x110eeef0110eeef0ULL
, },
67 { 0x4443bbbc4443bbbcULL
, 0x4443bbbc4443bbbcULL
, },
68 { 0xbd9fed0af683097cULL
, 0x84bc25eebd9fed0aULL
, },
69 { 0x97b2bda25ecfa130ULL
, 0xd09684be97b2bda2ULL
, },
70 { 0xaaa95556aaa95556ULL
, 0xaaa95556aaa95556ULL
, }, /* 24 */
71 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
72 { 0x71c638e471c638e4ULL
, 0x71c638e471c638e4ULL
, },
73 { 0x38e31c7238e31c72ULL
, 0x38e31c7238e31c72ULL
, },
74 { 0x8887777888877778ULL
, 0x8887777888877778ULL
, },
75 { 0x2221ddde2221dddeULL
, 0x2221ddde2221dddeULL
, },
76 { 0x5ecff6857b4184beULL
, 0x425e12f75ecff685ULL
, },
77 { 0x4bd95ed12f67d098ULL
, 0x684b425f4bd95ed1ULL
, },
78 { 0x9996666899966668ULL
, 0x9996666899966668ULL
, }, /* 32 */
79 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
80 { 0x110eeef0110eeef0ULL
, 0x110eeef0110eeef0ULL
, },
81 { 0x8887777888877778ULL
, 0x8887777888877778ULL
, },
82 { 0x47ab852047ab8520ULL
, 0x47ab852047ab8520ULL
, },
83 { 0x51eae14851eae148ULL
, 0x51eae14851eae148ULL
, },
84 { 0xe38cb60c27d071c8ULL
, 0x9f482d84e38cb60cULL
, },
85 { 0xb609b05c71c5f4a0ULL
, 0xfa4e38e4b609b05cULL
, },
86 { 0x6665999a6665999aULL
, 0x6665999a6665999aULL
, }, /* 40 */
87 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
88 { 0x4443bbbc4443bbbcULL
, 0x4443bbbc4443bbbcULL
, },
89 { 0x2221ddde2221dddeULL
, 0x2221ddde2221dddeULL
, },
90 { 0x51eae14851eae148ULL
, 0x51eae14851eae148ULL
, },
91 { 0x147ab852147ab852ULL
, 0x147ab852147ab852ULL
, },
92 { 0x38e32d8349f41c72ULL
, 0x27d20b6138e32d83ULL
, },
93 { 0x2d826c171c717d28ULL
, 0x3e938e392d826c17ULL
, },
94 { 0x1c6fe38f71c48e3aULL
, 0xc71a38e51c6fe38fULL
, }, /* 48 */
95 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
96 { 0xbd9fed0af683097cULL
, 0x84bc25eebd9fed0aULL
, },
97 { 0x5ecff6857b4184beULL
, 0x425e12f75ecff685ULL
, },
98 { 0xe38cb60c27d071c8ULL
, 0x9f482d84e38cb60cULL
, },
99 { 0x38e32d8349f41c72ULL
, 0x27d20b6138e32d83ULL
, },
100 { 0xd6e93c0d19474f04ULL
, 0x5ba64589d6e93c0dULL
, },
101 { 0x4586a782587d3f36ULL
, 0x6b73f35c4586a782ULL
, },
102 { 0xe38c1c738e3771c8ULL
, 0x38e1c71de38c1c73ULL
, }, /* 56 */
103 { 0x0000000000000000ULL
, 0x0000000000000000ULL
, },
104 { 0x97b2bda25ecfa130ULL
, 0xd09684be97b2bda2ULL
, },
105 { 0x4bd95ed12f67d098ULL
, 0x684b425f4bd95ed1ULL
, },
106 { 0xb609b05c71c5f4a0ULL
, 0xfa4e38e4b609b05cULL
, },
107 { 0x2d826c171c717d28ULL
, 0x3e938e392d826c17ULL
, },
108 { 0x4586a782587d3f36ULL
, 0x6b73f35c4586a782ULL
, },
109 { 0x9e0574f135ba3292ULL
, 0xcd6dd3c19e0574f1ULL
, },
110 { 0x18c3fe7422c25584ULL
, 0x16b6b9f57608cfa9ULL
, }, /* 64 */
111 { 0x867e6d904e841446ULL
, 0x0de4cfed4e2fdb15ULL
, },
112 { 0xf94f18bc4bc3d93eULL
, 0x1492568ac3a66499ULL
, },
113 { 0x4ff36c125a383042ULL
, 0x2fe23e4744196e36ULL
, },
114 { 0x867e6d904e841446ULL
, 0x0de4cfed4e2fdb15ULL
, },
115 { 0xf78e474db23f32a9ULL
, 0x8a26a8f51ca9cd91ULL
, },
116 { 0xa9bfb48aa4c2d0ddULL
, 0x94641c4e1a398e45ULL
, },
117 { 0x6e796f69cc7c8793ULL
, 0x6e879377578266beULL
, },
118 { 0xf94f18bc4bc3d93eULL
, 0x1492568ac3a66499ULL
, }, /* 72 */
119 { 0xa9bfb48aa4c2d0ddULL
, 0x94641c4e1a398e45ULL
, },
120 { 0xeb349888d2e11561ULL
, 0xa0e2f84177d142c9ULL
, },
121 { 0x5ad3b4e8bfaf139fULL
, 0x8076d98091fe5896ULL
, },
122 { 0x4ff36c125a383042ULL
, 0x2fe23e4744196e36ULL
, },
123 { 0x6e796f69cc7c8793ULL
, 0x6e879377578266beULL
, },
124 { 0x5ad3b4e8bfaf139fULL
, 0x8076d98091fe5896ULL
, },
125 { 0x33368b8aeab5d525ULL
, 0x97d9932138871904ULL
, },
128 reset_msa_registers();
130 gettimeofday(&start
, NULL
);
132 for (i
= 0; i
< PATTERN_INPUTS_SHORT_COUNT
; i
++) {
133 for (j
= 0; j
< PATTERN_INPUTS_SHORT_COUNT
; j
++) {
134 do_msa_DOTP_U_W(b128_pattern
[i
], b128_pattern
[j
],
135 b128_result
[PATTERN_INPUTS_SHORT_COUNT
* i
+ j
]);
139 for (i
= 0; i
< RANDOM_INPUTS_SHORT_COUNT
; i
++) {
140 for (j
= 0; j
< RANDOM_INPUTS_SHORT_COUNT
; j
++) {
141 do_msa_DOTP_U_W(b128_random
[i
], b128_random
[j
],
142 b128_result
[((PATTERN_INPUTS_SHORT_COUNT
) *
143 (PATTERN_INPUTS_SHORT_COUNT
)) +
144 RANDOM_INPUTS_SHORT_COUNT
* i
+ j
]);
148 gettimeofday(&end
, NULL
);
150 elapsed_time
= (end
.tv_sec
- start
.tv_sec
) * 1000.0;
151 elapsed_time
+= (end
.tv_usec
- start
.tv_usec
) / 1000.0;
153 ret
= check_results(instruction_name
, TEST_COUNT_TOTAL
, elapsed_time
,
154 &b128_result
[0][0], &b128_expect
[0][0]);